Back to skill

Security audit

BlockRandom

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious because its install command uses an unpinned GitHub revision that could run repository-controlled code during installation; scanners found no clear evidence of malicious intent.

Before installing, review the upstream source and consider using a vetted, fixed commit or verified package in an isolated, least-privilege environment. Skillspector's analysis was partial: it scanned the listed Markdown component, but some local references were unresolved, so the evidence is incomplete.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • YARA SignaturesMalware Match, Webshell Match, Cryptominer Match
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

YARA rule 'agent_skill_remote_bootstrap_execution': Remote script or code download followed by execution/bootstrap installation [agent_skills]

High
Category
YARA Match
Confidence
85% confidence
Finding

The installation command retrieves an unpinned GitHub revision, and pip installation can execute repository-controlled build code with the installing user's privileges. This is a supply-chain weakness rather than evidence of malicious bootstrap behavior: a compromised upstream repository could execute arbitrary code or install a tool that silently manipulates trial allocations. Installation is legitimate in this context, but the clinical-trial use increases the importance of software integrity, and the documented self-check does not independently establish trust.

Content

Scanner excerpt · SKILL.md (reported line 21)May include surrounding context.

o, block sizes, strata and a seed, it writes the full allocation list, a sealed enrolment list, per-stratum files, an audit log and a methods paragraph. Balance is exact at every completed block's end, and the same seed and parameters regenerate the lists byte-for-byte. Without a seed it refuses: an unseeded list cannot be reproduced (ICH E9 requires a reproducible schedule).

Install

bash
pip install git+https://github.com/RakTikoo/blockrandom

This provides the blockrandom command; verify an install with blockrandom check (block balance and byte-for-byte regeneration, PASS/FAIL).

Run

bash
blockrandom generate --trial-name "Example trial" --arms Treatment,Control --ratio 1:1 --blocks 4,6,8 --n-per-stratum 40 --strata "site=North,South" --seed-file seed.txt --out-dir trial-lists
  • --arms (at least two, all different), --ratio (one whole number per arm; 2:1 and three-arm ratios work), --blocks and --n-per-stratum are required and never guessed. Ev